Abstract
Analyzing how generalized quantum mechanics and g-bits reshape quantum probability and machine learning, revealing new patterns in uncertainty and prediction.
This paper analyses how generalized quantum mechanics and g-bits reshape quantum probability and machine learning, revealing new patterns in uncertainty and prediction.
Overview
The generalized bit replaces the two-outcome projective measurement with a family of effects whose probability assignments need not satisfy the Boolean constraints assumed by classical learning theory. The consequence is a modified notion of capacity, and with it a modified generalization bound.
Full derivations, the numerical experiments, and the complete reference list are in the PDF.